Assessment of geological risk in the Canary Islands: the GeoMEP project

The assessment of the losses which may be potentially caused by a natural disaster in a given scenario is an essential exercise in the insurance industry for the estimation of the necessary financial resources to deal with the disaster. Since it is not possible to use statistical data, due to the low frequency of causative catastrophic events, the development of modelling techniques was increasingly generalized after the hurricane Andrew (August 1992) as a way to forecast the damage. GeoMEP, a loss evaluation method for geologic risks, has been developed on the base of an agreement of collaboration between the Geological and Mining Institute of Spain (IGME) and the Consorcio de Compensación de Seguros (CCS). In the preliminary stage, GeoMEP has focused on Canary Islands as a pilot project, which can be extended to other regions of Spain.

The Canary Island archipelago is a virtually unparalleled natural laboratory, not just for its geological origin or climate but also for the natural processes that take place as a result of the two

The Canary Island archipelago is a virtually unparalleled natural laboratory, not just for its geological origin or climate but also for the natural processes that take place as a result of the two. A third element, high population density combined with tourist appeal, makes the islands an ideal site for studying all manner of geological hazards. Such research contributes, on the one hand, to meeting the population’s need for information on the natural dynamics of their immediate surrounds, and on the other, to the development of methodologies to assess possible outcomes.

Not all geological processes disrupt people’s everyday lives. Those that may indeed have adverse effects are called geological hazards. The Canary Islands are subject to many such hazards, including earthquakes, floods and volcanic eruptions. The foreseeable consequences of a geological hazard are known as “geological risks”.

In Spain the Consorcio de Compensación de Seguros (CCS, insurance compensation consortium), a public insurer, is entrusted with covering the losses caused by such risks. But if a hazardous event were to occur tomorrow, how could we determine what the economic implications would be? What liability should the CCS be prepared to confront? The Instituto Geológico y Minero de España (IGME, Spanish institute for geological surveys), in turn, has been conducting studies of natural risks since its founding in 1849, and has intensified that line of research considerably in recent decades. Prompted by a common interest in understanding natural risks, the IGME and the CCS have concluded an agreement to assess the geological risks in the Canary Islands ( GeoMEP projectt) (1) to estimate the resources required to provision for flooding, earthquake and volcanic risk.

As noted, risks are merely the foreseeable implications of the future occurrence of a hazardous event, while losses may entail more than that. Experience has shown that after natural episodes, cause and effect can rarely be univocally and unequivocally related without resorting to the study of case histories and cascade effects. Lava flows, ash rain and explosions may be observed during volcanic eruptions, for instance. Similarly, if a very shallow, high magnitude earthquake strikes in a steeply sloped area comprising unstable materials, it is unthinkable not to expect landslides. A flood consisting only of the submergence of normally dry land would be an exception, for these events generally go hand-in-hand with river bank and subsoil erosion, changes in the course of the river and sedimentation. Moreover, natural hazards do not always constitute a problem per se, but trigger other adverse situations: one recent example is the 2011 nuclear disaster in Japan, caused by a tsunami that was set off by an earthquake. Cascade effects are events in which an apparently minor occurrence gives rise to a second with feedback effects, such as the spread of fires after earthquakes.

While the use of the data on record, if any, to assess future consequences is tempting, it has proven to be a vain exercise, due both to the particular nature of (and cascade effects related to) specific events and to the speedy change to which an area’s economic, social and natural variables are subject. The hazard calculated for a series of past events is not correlated to the damage recorded. Nonetheless, when the variables space, time, exposure and vulnerability are fixed and only hazard varies, the resulting curve trends non-linearly upward, with thresholds, such as the “desk effect” in floods (2).

Solving the risk equation is no trivial matter, and the first obstacle to surmount is data availability. If a natural hazard database were in place, the task would be far easier, for this is the area where most of the limitations arise. From there on, as much complexity may be introduced into the study as required in each case, pondering matters such as direct or indirect damage, social or structural topology or cascade effects. The number of variables and considerations to be included in risk studies, as well as the demand for such studies and society’s capacity to contend with the uncertainties associated with the calculations involved, grow as science, technology and society in general move forward.

Certain features of the study of this first parameter are common to floods, earthquakes and volcanoes, while others are exclusive to each. The three shared pillars are: the application of mathematical models and numerical information, historic information, and geological information. Moreover, all these models and data are integrated into a single geographic information system (GIS).

The flooded area was determined using historical and geological data in addition to a numerical rainfall/runoff model and a model for determining the extent of flooding. The seismic hazard study of tectonic earthquakes (4) used a mathematical model to accommodate attenuation determined by distance and site parameters such as the types of materials involved and land relief. The volcanic hazard study took lava flow properties into consideration, along with the volume of earlier flows, applying a model that assesses the path and extent of future lava flows.

The seismic study conducted was associated with an occurrence in a specific area: a seismogenic geological structure, i.e., a structure able to generate earthquakes whose effects are propagated across the entire archipelago. As no other structure able to generate sizeable quakes has been identified, the seismic scenario is unique. In flooding and volcanism, however, the study of the area potentially affected involves deploying a model for randomising where rain may fall or a lava vent may open up. While in earthquakes only one possible event (a unique scenario) with a 0.2 % yearly chance of occurrence is identified, in flooding and volcanism the number of possible scenarios with the same likelihood is, if not infinite, very large and indeterminate because the events modelled, rainfall and lava emission, may take place anywhere on the map. In this study, several scenarios of the same likelihood were assessed with a view to ascertaining maximum loss.

Severity was another feature dealt with differently in each type of hazard. While in volcanism it was delimited by the path of the lava flow (spatial magnitude), in flooding, in addition to the potentially flood prone area, the depth at each point was factored into the model. For seismicity, severity was determined in terms of the ground acceleration across the archipelago, in turn converted into an intensity estimator (5).

Cadastre databases were used to study exposure. These databases include real estate appraisals divided into land value and building value. For flooding, seismicity and volcanism, the assumed insurable value was the value of the buildings (using a conversion factor to find the reconstruction value in the event of seismicity and flooding), plus a percentage to cover the value of the contents. While only the value of buildings was appraised for the seismic analysis, although including all the buildings on the archipelago (disregarding content losses), in the volcanism study both buildings and content located in the path of the lava flow were appraised for each scenario. In flooding, total exposure included the value of the movables in the ground and below grade storeys in buildings in the flooded area, and the buildings themselves only where a certain flood crest threshold was crossed (Table 1).

Cadastre database cleansing operations for GIS analysis included the elimination or correction, as appropriate, of topological errors such as overlaps (plots sharing the same space), duplications (perhaps the result of past revisions), open polygons, polygons with negligible areas or none at all, and gaps. Another question that had to be addressed was the discrepancy between alpha-numerical and graphic records, more than likely the outcome of separate updating of the graphic and alpha-numerical divisions of the database. Once the databases had been cleansed, the total value of all the buildings on each cadastral plot was found to reduce the number of elements to be studied and ensure the confidentiality of the information processed.

Cadastral plot vulnerability to lava flows was assumed to be 100 %; in other words, the damage to any plot affected by a lava flow would be equal to the value of the buildings standing on it and their contents. That assumption was an acknowledgement that infrastructures are not designed to bear either substantial horizontal loads or the temperatures reached by natural incandescent materials and that once the lava solidifies, the infrastructure is irrecoverable.

For seismicity, the reference adopted was the 1998 European Macroseismic Scale (EMS98), which grades expected earthquake damage by the theoretical intensity calculated. First the buildings were classified into one of the construction types set out in the EMS98 and, based on the intensity calculated for each building’s centroid, a percentage of expected damage was deduced from the EMS98.

A similar criterion was applied for flooding. Content losses were calculated on the grounds of the expected flood depth in each plot and known damage thresholds, while the value of buildings was only included for elevations at which structural damage could be expected.
 

Canary Island flood prone zones are normally narrow strips because the archipelago’s non-permanent and torrential hydrology (6) and steep slopes have formed deep riverbeds and narrow valleys. The essentially flat areas on the coast, however, or where valleys have widened, may give rise to larger flood zones with shallower depths and lower flow rates than in medium and high reaches.

The findings showed that ground and below grade storeys account for nearly 65 % of the total value, while the value of all the scenarios considered amounts to less than 30 % of the total. In other words, over 70 % of the total value of all the property in the archipelago lies outside the flood prone areas defined in the model. However, the model developed showed that not all the zones where the chance of yearly occurrence is 0.2 % can be flooded at the same time. Scenarios were consequently defined to foresee losses in ways consistent with a 0.2 % yearly likelihood. The least favourable of all the scenarios thus constructed were chosen and divided into two groups, depending on whether a single basin or a group of basins would be affected, as summarised below.

a) The worst-case scenarios for each island and one basin only are listed in the following table:

b ) Where the area of influence covered by the event was forecast to affect one or more basins, the results, relative to the size of the island and the mean area affected by the events simulated, were as follows:

Only direct effects, i.e., damage to buildings as the result of ground vibration, were taken into consideration when estimating seismic risk. The most vulnerable categories (where damage would be most severe, grades A and B on the EMS98 scale) were found to be less abundant than the least vulnerable (grades C and D). For that reason, most of the potential loss was identified in groups C and D, while the value of the loss in grades A and B was under 20 %. The geographic distribution of all the losses, i.e., summing the values for the four types of construction studied, is mapped in Figure 8. As the map shows, the variable with the heaviest weight in seismicity studies is population distribution, for the most severe damage was located in the largest towns and cities. Moreover, the islands farthest away from the epicentre defined in the study (sited between Tenerife and Gran Canaria Islands) happen to be the ones with the smallest populations. The total value of insured loss in the event of an earthquake was calculated to come to 353.37 million euros.

One particularity that sets volcanic hazard off from flood and earthquake hazards is that by changing the volume of lava, one and the same vent can give rise to different scenarios. More than that, if lava viscosity is altered, the same vent with the same volume of lava may also give rise to different scenarios. And to complicate matters further, any given area may be affected by more than one scenario generated by different vents. Moreover, building volcanic scenarios involves such large volumes of data and such high costs that, at this writing, scenarios are only available for the island of Tenerife. The results of the present study were consequently confined to the possible implications of an eruption on that island. Be it nonetheless said that of the seven islands, six are volcanically active (only La Gomera is regarded as inactive). Furthermore, only the part of the volcanoes above the ocean’s surface were studied, for no data are available for the submarine portion of the volcanic edifices, which is by far the largest. The volcanic structures that form the Canary Islands rise from four to nearly eight thousand metresfrom the ocean floor, although little more than the tip emerges.

Twelve of the fourteen scenarios selected to assess their possible consequences were found to exhibit significant risk, as shown in the table below.
 

Scenario	N. plots	Part insured 106 €
c526_20	12.052	5.504,43
c525_19	11.215	3.612,02
c520_17	7.501	1.253,92
c555_25	4.896	784,52
c829_1	838	522,58
c394_1	1.330	505,39
c395_3	665	406,84
c262_4	1.467	321,06
c930_3	294	189,63
c295_4b	769	82,58
c484_2	50	4,22
c494_2	2	0,04

The two scenarios found to cause the greatest damage were C525-19 and C526-20 (Figures 9 and 10). While both would affect Puerto de la Cruz, they differed in the location of the lava vents, which conditioned the area covered and the path of the lava flow.